# warning windows doublequotes

perl  -lne "BEGIN{$/=''} @arr = split /\s+/,$_;$key=join qq(\n),@arr[0
+..3]; push @{$hash{$key}},$arr[4];  $arr[4]=~/(\d+):/? $count{$key}+=
+$1:'';END{print map {qq($_ $count{$_} @{$hash{$_}}  \n\n)} keys %hash
+ }"    gen1.txt gen2.txt


@gi
AGAT
+
AAAA 2 1:1.txt 1:2.txt

@gi
CATT
+
AAAAA 3 3:1.txt

@gi
TACA
+
AAAA 2 2:2.txt



perl  -MO=Deparse -lne "BEGIN{$/=''} @arr = split /\s+/,$_;$key=join q
+q(\n),@arr[0..3]; push @{$hash{$key}},$arr[4];  $arr[4]=~/(\d+):/? $c
+ount{$key}+=$1:'';END{print map {qq($_ $count{$_} @{$hash{$_}}  \n\n)
+} keys %hash }" gen1.txt gen2.txt

BEGIN { $/ = ""; $\ = "\n"; }
LINE: while (defined($_ = <ARGV>)) {
    chomp $_;
    sub BEGIN {
        $/ = '';
    }
    @arr = split(/\s+/, $_, 0);
    $key = join("\n", @arr[0..3]);
    push @{$hash{$key};}, $arr[4];
    $arr[4] =~ /(\d+):/ ? $count{$key} += $1 : '???';
    sub END {
        print map({"$_ $count{$_} @{$hash{$_};}  \n\n";} keys %hash);
    }
    ;
}
-e syntax OK
[download]

UPDATE still silly but no need of a separate hash to hold the count and no need of an array to accumulate the strings.

                                                                      
+                       #fix missed { in {c}
perl  -lne "BEGIN{$/=''}@ar=split /\s+/,$_;$k=join qq(\n),@ar[0..3];$h
+{$k}{s}.=' '. $ar[4];$h{$k}{c}+=$1 if $ar[4]=~/(\d+)/;END{print map{q
+q($_ $h{$_}{c} $h{$_}{s}\n\n)} keys %h}"    gen1.txt gen2.txt

@gi
AGAT
+
AAAA 2  1:1.txt 1:2.txt

@gi
CATT
+
AAAAA 3  3:1.txt

@gi
TACA
+
AAAA 2  2:2.txt

# infact the dumped datastructure is easily intellegible:
# in the oneliner h is for hash
# s is for string
# c for count (ah! never use single char variable names!)

perl -MData::Dump -lne "BEGIN{$/=''}@ar=split /\s+/,$_;$k=join qq(\n),
+@ar[0..3];$h{$k}{s}.=' '.$ar[4];$h{$k}{c}+=$1 if $ar[4]=~/(\d+)/;END{
+dd %h}"    gen1.txt gen2.txt

(
  "\@gi\nAGAT\n+\nAAAA",
  { c => 2, s => " 1:1.txt 1:2.txt" },
  "\@gi\nCATT\n+\nAAAAA",
  { c => 3, s => " 3:1.txt" },
  "\@gi\nTACA\n+\nAAAA",
  { c => 2, s => " 2:2.txt" },
)


#oh well we dont need @ar at all, perl -a at the rescue:

perl  -lane "BEGIN{$/=''}$k=join qq(\n),@F[0..3];$h{$k}{s}.=' '.$F[4];
+$h{$k}{c}+=$1 if $F[4]=~/(\d+):/;END{print map{qq($_ $h{$_}{c} $h{$_}
+{s}\n\n)} keys %h}"    gen1.txt gen2.txt


# ok i've nothing better to do this night..

perl  -lane "BEGIN{$/=''}map{$h{$_}{s}.=' '.$F[4];$h{$_}{c}+=$1 if $F[
+4]=~/(\d+):/}join qq(\n),@F[0..3]; END{print map{qq($_ $h{$_}{c} $h{$
+_}{s}\n\n)} keys %h}"    gen1.txt gen2.txt
[download]

L*